Ask HN: How would you build a Grammerly clone for another language?

What is the process? What type of datasets are needed? What algorithms will be used?